In patients on chronic haemodialysis, because of a non-specific increase in serum ferritin, iron deficiency may be overlooked leading to failure of erythropoietin treatment. A reticulocyte haemglobin content < 26 pg and a percentage of hypochromic red cells > 2.5 have been proposed as markers of iron-deficient erythropoiesis in such subjects, but it is unclear which parameter is superior. We measured haematocrit, reticulocyte haemglobin content, ferritin and the percentage of hypochromic red cells over 10-150 days in 36 chronic haemodialysis patients in a university hospital. Transferrin saturation was also measured in a subset of 25 patients; iron deficiency was defined as a transferrin saturation < 15%. The diagnostic sensitivity and specificity of a reticulocyte haemoglobin content < 26 pg in detecting iron deficiency were 100% and 73% respectively, compared with 91% and 54% for a percentage of hypochromic red cells > 2.5. Paradoxical reticulocyte haemglobin concentrations occurred on follow-up in five patients receiving 4000 U erythropoietin per haemodialysis (HD). In three patients, reticulocyte haemglobin content exceeded 26 pg despite a persistent lack of iron. In a fourth, iron gluconate (62.5 mg i.v./HD) increased transferrin saturation to 27% and reduced the percentage of hypochromic red cells from 12 to 4, while reticulocyte haemglobin remained > 30 pg. In the final patient, iron gluconate increased transferrin saturation from 8 to 30% and reduced the percentage of hypochromic red cells from 40 to below 5, but reticulocyte haemglobin content remained < or = 26 pg throughout. The reticulocyte haemglobin content is superior to the percentage of hypochromic red cells in detecting iron deficiency in haemodialysis patients.